Question about TROCR model variations in terms of FLOPs and Inference time

microsoft / unilm

Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities

https://aka.ms/GeneralAI

MIT License

19.08k stars 2.43k forks source link

Question about TROCR model variations in terms of FLOPs and Inference time #1572

Open aliiafkari opened 2 weeks ago

aliiafkari commented 2 weeks ago

Greetings,

I have an inquiry regarding the TROCR paper. Could you please provide information on the number of FLOPs for each model variant (small, base, and large)? Additionally, I would appreciate details on the inference time for processing an image.

Thank you.